Development of an optimal experimental model for predicting rock mass rating based on tunneling quality index

Abstract

In initial phase of mining and civil projects, rock mass rating (RMR) and tunneling quality index (Q) classification systems are widely used. Because of specifying these values, it is possible to predict various rock mass characteristics, such as geomechanical parameters. Hence, many researchers developed experimental models so that the RMR is obtained from Q. The predicted RMR is close to the real value in some domains of Q, but for some domains it is deviated from the real value. Therefore, designers are confused to choose the optimal results of the models. The purpose of this paper is to develop an optimal model for determining RMR various domains of Q. Hence, firstly, the performance of the previous models is evaluated in six different domains of Q and then the best models for each domain are selected. Using the simple regression method, an optimal experimental model is developed based on the selected model in each domain. For this purpose, 214 datasets in the available literatures are used. Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) indicators are used to evaluate the performance of the models. Investigations show that the results of the developed model are more reliable than the other models for all of the domains. Therefore, the proposed models can be used as general models in all domains.